Skip to content
Longterm Wiki

Gemini 2.5 Flash

Google DeepMind

Gemini 2.5 Flash was previewed May 20, 2025 at Google I/O as a cost-efficient reasoning model. Featured hybrid reasoning with a toggleable thinking budget, 1M token context window, and multimodal capabilities. Priced at \$0.15/\$0.60 per million tokens. Designed to bring 2.5 Pro reasoning quality to Flash-tier pricing.

Developer
Google DeepMind
Released
2025-05-20
Context Window
1M tokens

Pricing

TypePrice per MTok
Input$0.15
Output$0.60

Benchmarks7

Knowledge

2 benchmarks
76%#8/15
50th percentile
86.6%#24/37
36th percentile

Reasoning

2 benchmarks
82.8%#5/34
87th percentile
64th percentile

Math

2 benchmarks
93.4%#9/31
73th percentile
72%#7/7
7th percentile

Coding

1 benchmark
42th percentile
Percentile among tested models:Top 25%50-75%25-50%Bottom 25%

Gemini Family5

ModelTierReleasedInput $/MTok
Gemini 2.5 Propro2025-03-25$1.25
Gemini 2.0 Flashflash2024-12-11$0.10
Gemini 1.5 Flashflash2024-05-24$0.07
Gemini 1.5 Propro2024-02-15$1.25
Gemini 1.0 Ultraultra2024-02-08

Details

Model FamilyGemini
Tierflash
Generation2.5
Release Date2025-05-20
Context Window1M tokens
Open WeightNo
Modalitytext, image, audio, video

Capabilities5

reasoningtool-usevisionaudio-inputlong-context

Sources3

Tags

geminigooglereasoningcost-efficient